Computer Vision News 48 Pioneers of Medical Robotics Another aspect of my research involves extracting information from imaging data, including semantic segmentation and depth estimation. Such information plays a vital role in various downstream tasks, such as 3D reconstruction, navigation, and surgical workflow analysis, and the quality of this extracted information could directly impact the performance of these tasks. During my PhD, I focused on semantic segmentation, a procedure of partitioning an image into multiple distinct regions representing different objects or classes. I developed algorithms capable of accurately segmenting endoscopic images using sparsely annotated data, by leveraging temporal information, increasing the training set with synthetic images, and enhancing feature representations. Moreover, I demonstrated how segmentation can benefit one of its downstream tasks, objective surgical skill assessment. Presently, I am extending my exploration to depth estimation and working on leveraging cross-modality information to enhance depth estimation along with semantic segmentation. While full automation of surgery is still far away, achieving smoother collaboration between surgeons and robots stands as the first milestone. My research will continue advancing perception algorithms and integrating multi-modal sensory data to create real-time, accurate "navigation maps" for surgery, empowering surgeons and robotic systems with a holistic understanding of the surgical scene, thereby facilitating informed and precise decision-making processes. Additionally, I will further integrate my perception results with tasks such as path planning, control, and manipulation, aiming to attain higher levels of autonomy and enhance treatment outcomes in surgery. The other Pioneer of Medical Robotics selected by the workshop.
RkJQdWJsaXNoZXIy NTc3NzU=